ROC (deeplearning4j-nn 0.9.1 API)

java.lang.Object
- org.deeplearning4j.eval.BaseEvaluation<ROC>
- - org.deeplearning4j.eval.ROC

All Implemented Interfaces:

Serializable, IEvaluation<ROC>
```
public class ROC
extends BaseEvaluation<ROC>
```
ROC (Receiver Operating Characteristic) for binary classifiers.
ROC has 2 modes of operation: (a) Thresholded (default, less memory)
(b) Exact (use numSteps == 0. May not scale to very large datasets)
Thresholded Is an approximate method, that (for large datasets) may use significantly less memory than exact.. Whereas exact implementations will automatically calculate the threshold points based on the data set to give a 'smoother' and more accurate ROC curve (or optimal cut points for diagnostic purposes), thresholded uses fixed steps of size 1.0 / thresholdSteps, as this allows easy implementation for batched and distributed evaluation scenarios (where the full data set is not available in memory on any one machine at once). Note that in some cases (very skewed probability predictions, for example) the threshold approach can be inaccurate, often underestimating the true area.
The data is assumed to be binary classification - nColumns == 1 (single binary output variable) or nColumns == 2 (probability distribution over 2 classes, with column 1 being values for 'positive' examples)

Author:

Alex Black

See Also:

Serialized Form

Nested Class Summary

Nested Classes
Modifier and Type Class and Description

static class ROC.CountsForThreshold

Nested Classes
Modifier and Type	Class and Description
`static class`	`ROC.CountsForThreshold`

Constructor Summary

Constructors
Constructor and Description
`ROC()`
`ROC(int thresholdSteps)`
`ROC(int thresholdSteps, boolean rocRemoveRedundantPts)`
`ROC(int thresholdSteps, boolean rocRemoveRedundantPts, int exactAllocBlockSize)`

Method Summary

All Methods Instance Methods Concrete Methods
Modifier and Type	Method and Description
`double`	`calculateAUC()` Calculate the AUROC - Area Under ROC Curve Utilizes trapezoidal integration internally
`double`	`calculateAUCPR()` Calculate the area under the precision/recall curve - aka AUCPR
`void`	`eval(org.nd4j.linalg.api.ndarray.INDArray labels, org.nd4j.linalg.api.ndarray.INDArray predictions)` Evaluate (collect statistics for) the given minibatch of data.
`PrecisionRecallCurve`	`getPrecisionRecallCurve()` Get the precision recall curve as array.
`protected org.nd4j.linalg.api.ndarray.INDArray`	`getProbAndLabelUsed()`
`RocCurve`	`getRocCurve()` Get the ROC curve, as a set of (threshold, falsePositive, truePositive) points
`void`	`merge(ROC other)` Merge this ROC instance with another.
`void`	`reset()`
`String`	`stats()`

Methods inherited from class org.deeplearning4j.eval.BaseEvaluation
equals, eval, eval, evalTimeSeries, evalTimeSeries, fromJson, fromYaml, toJson, toString, toYaml

Methods inherited from class java.lang.Object
clone, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

- Constructor Detail
  - ROC
```
public ROC()
```
  - ROC
```
public ROC(int thresholdSteps)
```
    Parameters:
    
    thresholdSteps - Number of threshold steps to use for the ROC calculation. If set to 0: use exact calculation
  - ROC
```
public ROC(int thresholdSteps,
           boolean rocRemoveRedundantPts)
```
    Parameters:
    
    thresholdSteps - Number of threshold steps to use for the ROC calculation. If set to 0: use exact calculation
    
    rocRemoveRedundantPts - Usually set to true. If true, remove any redundant points from ROC and P-R curves
  - ROC
```
public ROC(int thresholdSteps,
           boolean rocRemoveRedundantPts,
           int exactAllocBlockSize)
```
    Parameters:
    
    thresholdSteps - Number of threshold steps to use for the ROC calculation. If set to 0: use exact calculation
    
    rocRemoveRedundantPts - Usually set to true. If true, remove any redundant points from ROC and P-R curves
    
    exactAllocBlockSize - if using exact mode, the block size relocation. Users can likely use the default setting in almost all cases
- Method Detail
  - getProbAndLabelUsed
```
protected org.nd4j.linalg.api.ndarray.INDArray getProbAndLabelUsed()
```
  - reset
```
public void reset()
```
  - stats
```
public String stats()
```
    Returns:
  - eval
```
public void eval(org.nd4j.linalg.api.ndarray.INDArray labels,
                 org.nd4j.linalg.api.ndarray.INDArray predictions)
```
    Evaluate (collect statistics for) the given minibatch of data. For time series (3 dimensions) use BaseEvaluation.evalTimeSeries(INDArray, INDArray) or BaseEvaluation.evalTimeSeries(INDArray, INDArray, INDArray)
    
    Parameters:
    
    labels - Labels / true outcomes
    
    predictions - Predictions
  - getPrecisionRecallCurve
```
public PrecisionRecallCurve getPrecisionRecallCurve()
```
    Get the precision recall curve as array. return[0] = threshold array
    return[1] = precision array
    return[2] = recall array
    
    Returns:
  - getRocCurve
```
public RocCurve getRocCurve()
```
    Get the ROC curve, as a set of (threshold, falsePositive, truePositive) points
    
    Returns:
    
    ROC curve
  - calculateAUC
```
public double calculateAUC()
```
    Calculate the AUROC - Area Under ROC Curve
    Utilizes trapezoidal integration internally
    
    Returns:
    
    AUC
  - calculateAUCPR
```
public double calculateAUCPR()
```
    Calculate the area under the precision/recall curve - aka AUCPR
    
    Returns:
  - merge
```
public void merge(ROC other)
```
    Merge this ROC instance with another. This ROC instance is modified, by adding the stats from the other instance.
    
    Parameters:
    
    other - ROC instance to combine with this one

Class ROC

Nested Class Summary

Constructor Summary

Method Summary

Methods inherited from class org.deeplearning4j.eval.BaseEvaluation

Methods inherited from class java.lang.Object

Constructor Detail

ROC

ROC

ROC

ROC

Method Detail

getProbAndLabelUsed

reset

stats

eval

getPrecisionRecallCurve

getRocCurve

calculateAUC

calculateAUCPR

merge